Gene Feature Extraction Method Based on Cross-modal Nearest Neighbor Manifold Scatter
WANG Mengming
ZHANG Zhipeng
HOU Yakui
Abstract：In order to address the challenges posed by high-dimensional,small-sample,and noisy gene expression data in gene classification,this paper proposed the cross-modal nearest neighbor manifold scatter(CNNMS)method.The method utilized the nearest neighbor data based on the kernel method to further diminish the impact of class imbalance on classification accuracy.Additionally,leveraging the fact that the nearest neighbor mean is less influenced by outliers,CNNMS method mapped high-dimensional gene features to the kernel space.It defined the mean of the distance between all samples and their nearest neighbor samples as the nearest neighbor mean of the sample.This approach aimed to maintain clustering in the same feature class to the greatest extent in the multimodal nearest neighbor manifold dispersion subspace.The experimental results demonstrated that CNNMS method achieved a classification recognition rate of over 98%in lung cancer gene expression datasets and showed good classification recognition rate in gastric cancer gene expression datasets,and the method exhibited superior classification ability compared to other methods.CNNMS method proposed in the paper demonstrated a high recognition rate in gene classification research,bringing significant advancements to gene feature extraction.
Keywords：gene feature extractioncanonical correlation analysisdata dimensionality reductiongene classificationnearest neighbor scatterdiscrimination sensitivitycancer diagnosis
Publication Date：2024-03-20
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：5( 59-63 )